Risk, Return, and Income Mix at Commercial Banks:
Cross-Country Evidence
Rifat Gorener
Roosevelt University
Sungho Choi
Chonnam National University
The paper examines whether and how increased reliance on non-interest income affects the financial
performance of banks, as measured by stock market return data for publicly traded commercial banking
companies in 42 countries. In general, we find that non-interest income is associated with riskier stock
returns at commercial banking companies, due primarily to increased market, or systematic, risk. This
finding is new to the literature, and suggests that fee-based banking activities increase banks’ exposure to
the business cycle. In contrast, we find almost no evidence linking non-interest income to changes in the
total risk, interest rate risk, or idiosyncratic risk. Our results also suggest that the stock markets
efficiently price the increased risk associated with the non-interest income market. That is, after
controlling for cross-sectional differences in risk, market returns do not fluctuate with the mix of bank
income. This result offers a potential explanation for the initial conventional wisdom among industry
participants that expansion into non-interest activities would result in an improved risk–return trade-off
at commercial banks. Finally, we find that cross-country differences in regulatory practices, economic
conditions, and social institutions influence our main results in important ways, but on average our risk–
return results appear to be robust across countries.
INTRODUCTION
In the traditional sense, banks are intermediaries that add value and earn income based on the spread
between the interest paid on deposits and the interest received on loans. Market analysis of bank
performance, prudential regulation of banks, and theoretical models of bank “uniqueness” are all based
primarily on this framework of intermediation and spread income. However, non-interest or “fee-based”
activities account for a substantial and growing portion of the earnings in the banking sector. For
example, non-interest income comprised about 47 percent of the net revenue on average at the 25 largest
US banking companies in 2004, a rise from about 40 percent in the mid-1990s and about 35 percent in the
mid-1980s. Banking companies in other developed economies have experienced similar trends.1
The surge in non-interest income at commercial banks has come from two sources: the non-traditional
financial services that banks have only recently begun to provide and the traditional activities that
commercial banks have always provided but that are now produced and priced differently.2 During the
1990s, industry deregulation permitted commercial banks in many countries to expand into nontraditional product offerings, such as investment banking, merchant banking, insurance agency, and
Journal of Applied Business and Economics vol. 14(3) 2013
123
securities brokerage. Unlike traditional intermediation business, which generates an interest margin
between depositors and borrowers, these new banking activities primarily generate fee income.3 At
around the same time, the advances in information, communications, and financial technologies altered
the production and distribution of many traditional banking products; in the process, these products
became sources of non-interest income. For example, credit scoring and asset securitization have
transformed the production of consumer credit and home mortgages from a traditional portfolio lending
process in which banks earn mostly interest income into a transactions lending process in which banks
earn mostly non-interest income from loan origination and servicing fees. Similarly, some deposit
customers have demonstrated a willingness to pay higher fees (or accept lower interest rates on their
balances) for the convenience associated with widespread networks of branches, ATMs, and/or Internet
banking facilities.4
Initially, there was conventional wisdom among bankers, regulators, and industry analysts that
increased non-interest income would reduce the risk profile of commercial banks: less exposure to interest
rate movements and credit risk would stabilize bank revenues and diversification gains from a broader
business mix of fee-based activities would stabilize bank profits. However, the empirical literature has not
systematically confirmed these initial beliefs. While a handful of studies have found evidence linking
non-interest income to lower bank risk, a growing majority of studies are finding contradictory evidence.
In practice, the diversification gains appear to be limited (Stiroh, 2004) and/or tend to be consumed by
increased risk-taking in other areas (Demsetz & Strahan, 1997). Streams of non-interest revenue from
some activities appear to be more, rather than less, volatile than traditional loan-based revenue streams
and expose the bank to risk from increased operating leverage (DeYoung & Roland, 2001). Additionally,
at the average bank, expansion into non-interest activities appears to offer a poor risk–return trade-off
(DeYoung & Rice, 2004).
In this study, we examine whether and how increased reliance on non-interest income affects the
financial performance of banks, as measured by the stock market return data for 877 publicly traded
commercial banking companies in 42 countries between 1995 and 2002. By examining these relationships
in this broad, multi-national, financial markets framework, we advance the literature in at least three
important ways. First, we provide a global robustness test for previous studies, which have focused on
banks in single countries or small groups of similar countries. Second, most (though not all) previous
studies have been based on performance data from banks’ financial statements, and as such may reflect
the myriad economic misrepresentations embedded in accounting documents. Third, we can exploit the
substantial cross-country heterogeneity in our data to test how regulatory practices, economic conditions,
and social institutions affect these relationships.
Our main results are derived from a two-stage risk–return regression framework, which we estimate
multiple times using a variety of risk measures derived from a two-factor market model. In general, we
find that non-interest income is associated with riskier stock returns at commercial banking companies,
due primarily to increased market, or systematic, risk. The magnitude of this relationship is non-trivial: a
one-standard-deviation increase in cross-sectional non-interest income is associated with about a 6
percent increase in the market beta for the average bank in our sample. These findings are new to the
literature, and they suggest that fee-based banking activities (e.g., merger financing, loan origination fees,
brokerage commissions) increase banks’ exposure to the business cycle. In contrast, we find almost no
evidence linking non-interest income to changes in total risk (the variability of stock returns), interest rate
risk (a second factor in a market model), or idiosyncratic risk (the residual from a two-factor market
model).
Our estimation results suggest that, on average, the stock markets in these 42 countries efficiently
price the increased risk associated with the non-interest income market. That is, after controlling for
cross-sectional differences in risk, the market returns do not fluctuate with the mix of bank income.
However, when we re-estimate our models using accounting returns (ROA and ROE) rather than market
returns, we find evidence of a premium in accounting returns that is positively related to non-interest
income. This set of results is also new to the literature, and it offers a potential explanation for the initial
124
Journal of Applied Business and Economics vol. 14(3) 2013
conventional wisdom among industry participants that expansion into non-interest activities would result
in an improved risk–return trade-off at commercial banks.
Finally, we find that cross-country differences in regulatory practices, economic conditions, and
social institutions influence our main results in important ways, but on average our risk–return results
appear to be robust across countries. Because deregulation was important in removing the constraints that
prevented banks from participating in certain fee-based activities, reliance on non-interest income will
vary across banks based on country-specific regulations. In contrast, increased reliance on non-interest
income due to innovations in financial markets and information flows is less likely to vary across
countries, because technology allowing more efficient production of traditional banking products spreads
quickly. In general, we find that tight controls of the banking system designed to reduce risk (e.g., entry
restrictions, explicit deposit insurance, activity restrictions, ownership restrictions, competition policy) are
associated with less risky non-interest activities.
Our findings have implications for regulation. Activities that generate non-interest income can be less
transparent than traditional banking activities, and thus more difficult to monitor for safety and soundness.
For example, reductions in equity capital caused by credit losses are obvious and easily identifiable (i.e.,
the progression of loan delinquency, loan classification, loan provisioning, and loan charge-off), but
shocks to non-interest income affect equity capital more subtly through ex post reductions in retained
earnings, which for most established firms are the primary source of capital. Activities that generate
volatile earnings streams make this source of capital riskier, and this volatility is exacerbated by financial
leverage. Under the current regulatory capital rules, banks are not required to hold capital against most
fee-generating activities.5 Our findings invite a discussion about whether the risk associated with certain
non-interest activities is large enough and systematic enough to merit a required capital charge in future
versions of these capital regulations. Our findings also have implications for investors. If the additional
risk associated with non-interest income is predominantly systematic risk (as we find here), then investors
will require higher expected returns to accept this non-diversifiable risk in their portfolios.
The rest of paper is set out as follows. In Section 2, we discuss the literature that investigates the
relation between non-interest income and bank profitability and risk. Section 3 describes our data set and
presents the methodology employed in our paper. In Section 4, we report our empirical results. Section 5
examines the effect of cross-country institutional difference on banks. Then, in Section 6, we conclude
our analysis with a brief summary of our main findings and an assessment of their implications.
LITERATURE REVIEW
The impact of non-interest income and non-traditional activities on commercial bank performance has
been the subject of academic and regulatory study for three decades. Most of these studies were
performed prior to the large expansion of permissible commercial bank activities made possible by the
major deregulatory acts, e.g., the Gramm–Leach–Bliley (GLB) Act of 1999, allowing US financial
holding companies to participate in both banking and non-banking activities, and the Second Banking Coordination Directive of 1989 (implemented in 1993–94), permitting universal banks to expand anywhere
within the EU regardless of the host country restrictions on product powers. These early studies were
limited to examining the small number of non-interest activities permissible at the time, and they
produced a mixture of interesting although somewhat contradictory results. In general, the studies found
that diversifying into non-traditional financial activities could potentially reduce risk, but in practice the
post-diversification changes in risk followed no clear patterns, either increasing, decreasing, or remaining
unchanged across the various studies. Moreover, the risk reductions that were achievable in practice
tended to diminish quickly, and in some instances were reversed as banks ramped up their risk-taking in
other areas, such as greater financial leverage. More recent studies—performed after the implementation
of the major deregulatory acts, and also in the wake of the financial and technical innovations that
changed the production processes for generating both interest and non-interest income—have been able to
examine a much broader set of non-interest activities. These later studies have generally concluded that
expansion into fee-based financial activities has increased the riskiness of banking companies.
Journal of Applied Business and Economics vol. 14(3) 2013
125
The earliest group of studies provided suggestive evidence that banks could potentially reduce their
riskiness by diversifying into non-banking activities. Heggestad (1975), Johnson and Meinster (1974),
Litan (1985), and Wall and Eisenbeis (1984) used industry-level IRS data from the 1950s, 1960s, and
1970s to compare the aggregate earnings stream of the banking industry with the aggregate earnings
streams of other financial industries (e.g., securities firms, insurance companies, real estate brokers,
leasing companies, thrift institutions). While the results of these studies did not always agree across
industries, a common thread ran through the studies: over long periods of time, banking industry earnings
and non-banking financial industry earnings were quite uncorrelated with each other, and in extreme
cases these correlations were close to zero or even negative. This basic result suggested that if banks were
allowed to add some non-banking financial products to their traditional mix of banking services, the
resulting portfolio diversification effects could potentially increase banks’ expected returns without
increasing their riskiness (or, equivalently, reduce banks’ riskiness without reducing their expected
returns).
Most studies have used firm-level data rather than industry averages, and have found mixed results.
One set of these studies concluded that diversification into non-banking activities increased the riskiness
of banks. Boyd and Graham (1986) examined large bank holding companies (BHCs) that diversified into
non-banking activities during the 1970s and concluded that, in the absence of strict regulatory oversight
and control, expansion into these activities can increase the risk of failure. Sinkey and Nash (1993) found
that commercial banks that specialized in credit card lending (an often-securitized type of lending that
generates substantial fee income) generated higher and more volatile accounting returns, and had higher
probabilities of insolvency, than commercial banks with traditional product mixes during the 1980s.
Demsetz and Strahan (1997) studied the stock returns of BHCs and found that greater diversification
across product lines did not necessarily reduce risk because the diversifying BHCs tended to shift to
riskier mixes of activities and hold less equity. Roland (1997) discovered that abnormal returns from feebased activities were less persistent (more short-lived or volatile) than abnormal returns from lending and
deposit-taking at large BHCs. Kwan (1998) found that the accounting returns of Section 20 securities
affiliates tended to be more volatile, but not necessarily higher, than the accounting returns of their
commercial banking affiliates.
On the contrary, other firm-level studies found that diversifying into non-banking activities reduced
the bank risk, although these gains tended to be limited in size, scope, or practice. Boyd et al. (1980)
measured the correlations between accounting returns at the bank and non-bank affiliates of BHCs during
the 1970s, and found that the potential for risk reduction was exhausted at relatively low levels of nonbanking activities. Eisenbeis, Harris, and Lakonishok (1984) found positive abnormal stock returns
associated with the formation of one-bank holding companies between 1968 and 1970, a brief time period
during which these firms were permitted to engage in a wide variety of non-banking activities. Kwast
(1989) examined the accounting returns of the securities and non-securities activities of commercial
banking companies between 1976 and 1985, and found limited potential for risk reduction by diversifying
into securities activities. Brewer (1989) reached similar results using the market returns of US bank
holding companies. Gallo, Apilado, and Kolari (1996) ascertained that high levels of mutual fund activity
were associated with increased profitability, but only slightly moderated risk levels, at large BHCs
between 1987 and 1994. Rogers and Sinkey (1999) found that non-traditional activities were associated
with larger size, smaller interest margins, less core deposit funding, and less risk.
Because the set of permissible non-banking activities was substantially constrained prior to the late
1990s and early 2000s, an alternative research approach examined the return streams of unrelated banking
firms and non-banking financial firms, and then calculated the hypothetical reduction in earnings
variability based on the covariances of those earnings streams. Rosen et al. (1989) found minimal
financial benefits from the hypothetical diversification of banking companies into real estate activities.
Wall et al. (1993) constructed synthetic portfolios based on the accounting returns of banks and nonbanking financial firms, and concluded that banks would have experienced higher returns and lower risk
had they been able to diversify into small amounts of insurance, mutual fund, securities brokerage, or real
estate activities during the 1980s. Using both accounting data and market data from the 1970s and 1980s,
126
Journal of Applied Business and Economics vol. 14(3) 2013
Boyd, Graham, and Hewitt (1993) concluded that BHCs could have reduced their riskiness by merging
with life insurance or property/casualty insurance firms, but would likely have increased their riskiness by
merging with securities or real estate firms. Laderman (1998) applied a similar synthetic merger approach
to data from the 1980s and 1990s, and concluded that BHCs could have reduced the volatility of their
accounting returns by offering “modest to relatively substantial amounts” of life insurance or casualty
insurance underwriting. Allen and Jagtiani (2000) used stock market data to construct return streams for
synthetic “universal banks” consisting of a commercial banking company, a securities firm, and an
insurance company, and found that the exposure to market risk increased with the addition of these nonbanking activities. While these studies yielded provocative results, this research approach is limited
because (by construction and by necessity) it cannot capture the positive and/or negative synergies in
production, marketing, or organizational control from combining banking activities with non-banking
activities.
Despite the very mixed evidence produced by the academic research outlined above, by the late 1990s
there was still conventional wisdom among industry participants that fee-based activities had a stabilizing
effect on bank income and that diversifying into non-banking activities reduced bank risk. This
conventional wisdom was documented by DeYoung and Roland (2001), who also provided three new
conceptual arguments for why non-interest income may be less stable than income from traditional
banking activities. First, the revenue from traditional lending activities may be relatively stable over time
because switching costs and information costs make it costly for either borrowers or lenders to walk away
from a lending relationship; in contrast, the revenue from some fee-based activities may be relatively
unstable because banks face a high level of competitive rivalry, low information costs, and fluctuating
demand in a number of these product markets (e.g., investment banking, retail brokerage, securitized
mortgage refinance). Second, fee-based services can require the bank to increase its ratio of fixed-tovariable expenses (e.g., skilled labor, retail sales space); the higher operating leverage that results
increases the sensitivity of the bank earnings to fluctuations in the bank revenues. Third, although banks
internally allocate some capital to their fee-based lines of business, there is no regulatory capital
requirement for most of these activities, which suggests a higher degree of financial leverage—and thus
higher earnings volatility—for these activities. Using quarterly data from US commercial banks between
1988 and 1995, DeYoung and Roland (2001) showed that non-interest income (from non-deposit-related
activities) was associated with more volatile revenue streams, a higher degree of total leverage, and more
volatile earnings streams, when compared with interest-based lending activities.
Similarly, the studies of US banking companies that followed also contradicted the conventional
industry wisdom. Cooper, Jackson, and Patterson (2003) found that non-interest income was associated
with lower risk-adjusted stock returns at bank holding companies. Stiroh (2004a) found little riskreduction benefit for small banking companies that engaged in both traditional and non-traditional
banking activities, although he found substantial diversification benefits for small banks that diversified
within traditional or non-traditional areas. In another study that included data from both large and small
banks, Stiroh (2004b) found reductions in earnings volatility at banks with large amounts of non-interest
income, but concluded that these were due to the reduced volatility of net interest income at those banks,
and were not due to diversification gains. DeYoung and Rice (2004) concluded that expansion into noninterest activities appeared to offer a poor risk–return trade-off. Studies using non-US data remain rare,
and to date have found mixed results. Esho, Kofman, and Sharpe (2005) found that increased reliance on
fee-based activities was associated with increased risk at Australian credit unions. Like the US studies,
Smith, Staikouras, and Wood (2003) found that non-interest income exhibited more volatility than net
interest income over time; the negative correlations between these two income streams led them to
conclude that diversification effects exist. Overall, these more recent studies have benefited from the
availability of more detailed information about the composition of non-interest-based activities and, since
they used more recent data either wholly or partially drawn from post-deregulation markets, as such
reflected less constrained behavior and performance of banking companies.
Journal of Applied Business and Economics vol. 14(3) 2013
127
METHODOLOGY
Our primary objective is to determine whether and how non-interest income affects the riskiness of
banking companies, and whether the market returns of these companies are sensitive to any such changes
in risk. To answer these questions, we first need to generate bank-level measures of risk (RISK) and
return (RETURN). We use the following two-factor market model to generate bank-level measures of
RISK:
Rit = αi + βi MRit + γi INTit + εit
(1)
where Rit is the return on the stock of bank i during period t, MRit is the return on a stock market index in
bank i’s country during period t, INTit is the change in the market interest rate on a benchmark long-term
government bond in bank i’s country during period t, and the error term εit is assumed to be distributed
symmetrically with mean zero. We estimate (1) separately for each bank i in each year of our 1995–2002
database, using weekly data (i.e., t = one week) and ordinary least squares techniques with robust errors.
We use four different RISK measures in our tests, each of which is derived from equation (1).
Systematic or market risk (MKTRISK) is the estimated value of βi. Interest rate risk (IRRISK) is the
estimated value of γi. Idiosyncratic risk (IDIORISK) is the standard deviation of the estimated residual
terms εit. Total risk (TOTRISK) is the standard deviation of Rit. Our primary RETURN measure is the
annualized average of the weekly returns Rit. In alternative tests, we use two accounting return measures,
the annual return on assets (ROAi) and the annual return on equity (ROEi).
We estimate the relationships between non-interest income and banking company performance in the
following system of RISK–RETURN equations:
RISKit = μ + λNIIit + ψYit + τΣt=1,TTt + πit
(2)
RETURNit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
(3)
which we estimate using two-stage least squares estimation techniques and an unbalanced panel of annual
data from 1995 through 2002.6 NII equals the ratio of non-interest income to total assets at bank i in year
t. We use alternative definitions of NII in robustness tests. Thus, the main test statistics are the estimated
coefficients λ, φ, and δ. Yit and Zit are vectors of control variables (described below). T is a vector of time
fixed-effects dummies. We also use bank fixed effects and country fixed effects in robustness tests. The
error terms πit and ηit are assumed to be distributed symmetrically with mean zero.
The coefficient λ has a straightforward interpretation: it is the sensitivity of a bank’s riskiness to
increases in its non-interest income. As the previous literature has found evidence of both positive and
negative associations between non-interest income and risk, we have no a priori expectations about the
sign of λ. However, the interpretation of λ will depend on how RISK is defined. If RISK = TOTRISK,
then we are testing whether non-interest income increases the risk exposure of undiversified stakeholders,
such as bank employees, bank managers, or bank regulators. However, if RISK = IDIORISK, then we are
testing whether the risk associated with non-interest income is idiosyncratic to bank i, and hence is of no
concern to a diversified investor. Conversely, if RISK = MKTRISK, then we are testing whether noninterest income increases banks’ exposure to the ups and downs of the stock market and the economy in
general, creating risk that cannot be avoided by diversified shareholders. Finally, if RISK = IRRISK, then
we are testing whether non-interest income reduces banks’ exposure to movements in interest rates, part
of the conventional wisdom initially posited by bankers and industry analysts.
The coefficient δ also has a straightforward interpretation: it measures the marginal risk–return tradeoff required by investors in banking company stock. Obviously, we expect this effect to be positive in an
efficient market. Because this trade-off may not be linear, we also estimate alternative regression
specifications that include both linear and quadratic RISK terms. The coefficient φ has a more subtle
128
Journal of Applied Business and Economics vol. 14(3) 2013
interpretation: it is the sensitivity of stock returns to increases in non-interest income after controlling for
any change in the risk–return trade-off (i.e., δRISK) caused by the increase in non-interest income. A zero
estimated value for φ would be consistent with an efficient stock market in which the risks associated with
non-interest activities are priced just as efficiently as the risks associated with more traditional, interestbased banking activities. However, it is possible that the risks associated with relatively new banking
activities are not yet well understood by investors, and hence may be poorly priced.7 If the market
systematically underprices (overprices) the risk associated with non-interest activities, then we expect a
positive (negative) φ coefficient.
Given the large number (42) of nations in our data, it is not possible to observe the various sources of
non-interest income and how they may differ across banks. Our non-interest income variable NII
combines all the non-interest income earned by a bank—fees, commissions, trading gains/losses, etc.—
from all its lines of business. Because demand schedules and production functions can vary greatly across
these different lines of business, the unobserved composition of NII may drive some of our findings. For
example, if we find that NII is positively associated with MKTRISK, then we might infer that NII is
weighted toward activities that are especially sensitive to the business cycle, such as investment banking,
merger finance, or retail brokerage. Similarly, if we find that NII is positively associated with IRRISK,
then we might infer that NII is weighted toward activities that are especially sensitive to interest rate
fluctuations, such as the fee income earned via loan origination, securitization, and servicing. Not having
access to the line-of-business composition of non-interest income also limits our ability to test for
diversification effects. At best, we can draw inferences about whether combining non-interest income
with all the other banking activities creates diversification gains. For instance, a finding that NII is
positively associated with MKTRISK, but negatively associated or neutral with respect to TOTRISK,
would imply that an increase in non-interest income generates gains from diversification within the
average bank. These gains could be due to cost synergies (input sharing), revenue synergies (e.g., crossselling), or less than perfect covariation of the cash flows between interest-based and non-interest-based
activities.
The variable definitions, data sources, and summary statistics for all of the regression variables used
to estimate (1), (2), and (3) are displayed in Table 1. Our primary data source was the Fitch-IBCA
Bankscope database, from which we drew our initial data set of 14,404 financial institutions and banks
from 46 different countries observed annually from 1995 through 2002. From this large initial sample, we
retained only publicly traded bank holding companies (BHCs), commercial banks, and savings banks for
which we had full information.8 This resulted in a final data set of 877 banking companies from 42
countries. Table 1 displays some descriptive information for our bank data set. The accounting data for
banks come from the Fitch-IBCA Bankscope database and the market data, including individual stock
return data, interest rate data, and market index, for each country come from Datastream. The mean NII is
1.85 percent with a range from -6.27 percent to 39.26 percent, which indicates the wide variation among
our sample banks. For the US studies, DeYoung and Rice (2004) reported that the aggregate non-interest
income was 2.39 percent for the year 2001 and differed greatly between larger banks and smaller banks.
Smith et al. (2003) reported that in EU countries, the average non-interest income increased from 0.88
percent of the total assets in 1994 to 1.09 percent in 1998. The difference among studies is mainly caused
by the different samples under study. The mean TOTRISK is 2.28 percent, with a range from 0.08 to
39.45 percent. Our sample banks on average are less risky than the overall market of their country: the
mean market risk is 0.45, which is substantially smaller than the market beta of 1. The mean of the
interest rate risk and bank-specific risk is -0.3268 and 0.0207, respectively. Finally, the average stock
return of banks is 3.46 percent per annum, with wide variation among banks.
Control Variables
We include vectors of control variables in equations (2) and (3), denoted as Y in the RISK regression
and Z in the RETURN regression. These variables are included to specify each equation better and to
identify the system of equations. In addition, we include a vector of fixed time effects T in each
regression to absorb the average year-to-year variation in risk and return in the banking sector.
Journal of Applied Business and Economics vol. 14(3) 2013
129
Six of the control variables appear in both the RISK equation (2) and the RETURN equation (3).
lnASSETS is the natural logarithm of assets for bank i. Size may be risk-reducing because it increases the
potential for diversification and gives banks greater access to sophisticated risk-management tools;
however, the very largest banks may have greater systematic risk since their earnings are likely to be
closely aligned with the general economy. Holding risk constant (recall that RISK appears on the righthand side of the RETURN regression), the returns will increase (decrease) with the bank size if the
average bank experiences scale economies (diseconomies); if the bank size is associated with a highvolume, low-margin, transactions banking strategy, then the returns may decline with the bank size
(DeYoung, Hunter, & Udell, 2004); and there may be higher returns if the market prices-in a “too big to
fail” premium for the very largest banks. LOANS is the ratio of total loans to total assets for bank i. A
high loan-to-asset ratio may be risk-increasing because loans expose the bank to greater credit risk and
greater market risk than other assets do; however, this ratio may signal an efficiently run bank, and hence
less risk due to effective risk management. Holding risk constant, a high loan-to-asset ratio is likely to
generate higher returns. LIQUID is the ratio of liquid assets to total assets for bank i. A store of liquid
assets may represent a risk cushion that allows a bank to take on more risk, or it could signal risk-averse
management, in which case it will be associated with lower levels of risk. Holding risk constant, a large
store of low-yielding liquid assets is an opportunity cost and hence will be likely to be associated with
lower returns. LOSSPROV is the ratio of provisions for loan losses to total assets for bank i. Assuming
that this ratio indicates a bank’s expected credit risk, it should be associated with higher total risk, market
risk, and idiosyncratic risk. Holding risk constant, high provisioning should reduce the returns.
EXPLICIT is a dummy variable indicating that bank i operates in a country with explicit deposit
insurance. If explicit deposit insurance schemes increase the moral hazard incentives for bank managers,
banks may take more risks on average; however, if market investors interpret this policy as an implicit
government guarantee, the market risk may decline. Depositors can discipline banks by demanding higher
interest rates or withdrawing their deposit. However, the explicit deposit insurance reduces depositors’
incentives to monitor a bank. Therefore, the explicit deposit insurance lowers banks’ interest expenses
and makes interest payments less sensitive to bank risk. Therefore, we expect that EXPLICIT has a
positive impact on returns. CR3 is the three-bank concentration ratio in bank i’s country. In the absence of
competitive rivalry (high CR3), managers may choose the “quiet life,” in which case the risk will decline,
and/or they may slacken off and run the bank poorly, in which case the earnings could grow more
volatile. Holding risk constant, high market concentration will be associated with higher returns due to
less competitive rivalry.
There are five control variables that appear only in the RISK equation (2). CAPITAL is the ratio of
market-value equity to book-value assets for bank i. A large store of market-value equity (relative to the
size of the bank) indicates a vote of confidence from investors that allows the bank to take on more risk
without immediate penalty. RESTRICT is an index constructed by Barth, Caprio, and Levine (2001) that
increases with the restrictions on permissible bank activities (e.g., securities, insurance, real estate) in
bank i’s country. To the extent that these restrictions prevent banks from achieving their most efficient (or
most preferred) risk–return trade-off, this index could be either positively or negatively related to risk.
CORPGOV is an index constructed by Kaufman, Kraay, and Zoido-Lobaton (the KKZ index,9 1999 and
2005) that increases as the governance environment in bank i’s country improves. We expect this index to
be negatively related to risk, because poor corporate governance is typically associated with higher risktaking by managers. BANKFREE10 is an index that increases as restrictions on the banking and finance
decline (easier entry, less state ownership, less government credit allocation, etc.) in bank i’s country. The
sign of this variable is ambiguous: while greater bank freedom allows banks to engage in risky activities,
it also removes the impediments to achieving an efficient risk–return trade-off. MULTSUPS is a dummy
variable that indicates that bank i’s country has more than one bank supervisory body. If coordination
between supervisors is difficult, or if multiple chartering authorities precipitate a “race to the bottom,”
then this variable will be associated with greater risk.
There are two control variables that appear only in the RETURN equation (3). The coefficients for
these variables should be interpreted holding risk constant. STATE and FOREIGN are dummy variables
130
Journal of Applied Business and Economics vol. 14(3) 2013
and measure the ownership of bank i by either the state or foreign, respectively. The expected coefficient
sign for STATE would depend on whether bank i is a private bank expecting greater returns because
state-owned rival banks provide poor competition or a state bank expecting lower returns. A similar story
holds for foreign ownership.
BASIC RESULTS
The results from the full-sample estimations of equations (2) and (3) are displayed, respectively, in
Tables 2 and 3. There are three main findings in these regressions. First, non-interest income alters the
composition of risk at the average bank in our sample, but neither increases nor decreases the overall risk.
Second, we find positive associations in the data between the various risk measures and market returns, an
indication that investors are pricing bank risk in these 42 countries. Third, after controlling for risk, we
find no statistical relationship between non-interest income and market returns—that is, there is no
premium or discount associated with non-interest activities on average.
In Table 2, we find a positive and statistically significant association between non-interest income and
bank risk only when risk is defined as MKTRISK. Based on our estimates, a one-standard-deviation
increase in NII is associated with a 6 percent increase in market risk at the means of the data. There is no
evidence linking non-interest income to total risk, interest rate risk, or idiosyncratic risk. These results
imply the existence of gains from diversification: greater reliance on non-interest activities makes the
average bank in our data more sensitive to general market volatility (systematic risk) without increasing
its overall return volatility.
In Table 3, we find positive and statistically significant risk–return relationships, evidence that
investors require higher returns in exchange for accepting higher (diversifiable and non-diversifiable)
risks. Based on our estimates, one-standard-deviation increases in TOTRISK, MKTRISK, and IDIORISK
are associated with 82 percent, 94 percent, and 80 percent increases in RETURN, respectively, at the
means of the data. Given this robust evidence of risk pricing in the data, we conclude that the lack of
statistical significance for IRRISK in column [c] is likely to be due to the difficulties associated with
measuring interest rate risk. The coefficient for NII is statistically non-significant in all four regressions,
consistent with efficient markets that bid away any excess risk-adjusted returns associated with the
product mix.
Control Variables
The bank’s total assets might be the most obvious factor related to the level of bank risk. In Table 1,
the sign of LogAssets is negative and significant in three regression models, except the market risk model,
which indicates that larger banks are relatively more diversified, which lowers their total, bank-specific,
and interest risk. However, as banks’ size grows, they involve more non-interest income activities, which
have more comovements with the market, and thus their systematic market risk increases. The bank size
has a negative effect on the stock return controlling for the risk–return trade-off. This result indicates that
the average bank experiences scale diseconomies or is associated with a high-volume, low-margin,
transactions banking strategy.
LOANS is negative and significantly associated with all four risk measures. The results indicate that a
high loan-to-asset ratio may signal an efficiently run bank, and hence less risk due to effective risk
management. Holding risk constant, however, we find that a high loan-to-asset ratio is likely to generate
lower returns. The measure of financial leverage (CAPITAL) is positive and significantly associated
with total risk as well as market risk measures. The result is consistent with our expectation: a large store
of market-value equity indicates a vote of confidence from investors that allows the bank to take on more
risk without immediate penalty. The measure of asset liquidity (LIQUID) is positively and significantly
related to most risk measures and this effect might be due to the fact that liquid assets are shifty assets that
are more difficult to monitor and can be moved quickly into other investments, meaning that as banks
increase their liquid assets, the risk increases. The liquidity reduces the profitability significantly. The
ratio of a provision for loan losses to total assets (LOSSPROV) in our regression to control for the credit
Journal of Applied Business and Economics vol. 14(3) 2013
131
risk is positively related to all the risk measures. Obviously, LOSSPROV reduces the performance of a
bank significantly.
The country-specific variables are significantly associated with the risk of banks as well as the
performance of banks. The dummy variable for a market-based economy (MARKET) is positive and
weakly significant, indicating that banks in a market-based country have higher risk than banks in a bankbased country. This is mainly due to the limited positive role of banks in a market-based country. The
existence of explicit deposit insurance (EXPLICIT) reduces the risk of banks. This result is not consistent
with the moral hazard of bank managers. The result indicates that market investors interpret this policy as
an implicit government guarantee, resulting in lower risk. However, we do not find any significant effect
on the stock return. CORPGOV is highly significant and negative in risk regressions. This suggests that
better governance reduces the risk of banks significantly. Better governance is associated with lower risktaking behavior of managers in general. BANKFREE is negative and significantly associated with bank
risk. This indicates that if a bank is more free and open, then it can achieve an efficient risk–return tradeoff. The banking concentration increases the bank’s market risk significantly, while it also increases its
profitability. This is consistent with the “too-big-to-fail” argument. Multiple existences of bank
supervisory bodies increase banks’ risk in general. The state ownership of banks significantly reduces the
performance of banks.
Non-Linear Risk–Return Trade-Off
It is unlikely that the risk–return trade-offs for non-interest banking activities and interest-based
banking activities will be the same. Non-interest income derived from traditional banking activities, such
as fees charged to core depositors, is likely to be relatively stable, while non-interest income derived from
less traditional activities, such as investment banking, securities brokerage, or mortgage
origination/securitization, may fluctuate substantially with the local, regional, or international business
conditions (DeYoung & Roland, 2001). We cannot test this proposition directly because, as stated above,
our non-interest income data are not reported by lines of business; however, if we reasonably assume that
banks with especially large volumes of non-interest income tend to have less traditional business mixes,
then we can test this proposition indirectly. Table 4 shows the partial results from alternative
specifications of equation (3) that include the interaction variable RISK*NII, and reports the partial
derivatives with respect to both RISK and NII. The coefficients of the RISK*NII term offer weak
evidence that the market requires a higher return by banks with especially high levels of non-interest
income: these coefficients are positive (negative for IRRISK) but not statistically significant. We conduct
two additional tests in Table 4. First, we calculate a partial derivative of return with respective to NII and
RISK and test whether the partial derivatives are statistically different from zero. Generally, we find that
the partial derivatives are significantly different from zero, which suggests that both NII and Risk have a
significant effect on the returns of banks. Second, we conduct the joint significance test to determine the
joint significance of NII and RISK*NII as well as the joint significance of RISK and RISK*NII. While
we find that NII and RISK*NII are weakly jointly significant, RISK and RISK*NII are always jointly
significant. Overall, the findings suggest weak evidence that the market requires a higher return from
banks with especially high levels of non-interest income.
Accounting Returns
Strategic investment and product mix decisions at all firms, including banking companies, are often
made and evaluated based on accounting information rather than market returns. Perhaps the main reason
for this is that, while it is difficult to isolate the impact of a given business decision or investment project
on a firm’s stock price, the project’s accounting-based rate of return can be calculated by tracing the
impact of that project on the revenue and expense lines in the firm’s income statement. It is also likely
that managers make investment and product mix decisions based on the probable impact of those
decisions on their firms’ financial statements, because their promotions and compensation (and this is
especially true for middle-level and quasi-upper-level managers) are heavily influenced by easily
quantifiable information in financial statements. Thus, having found no statistical relationship between
132
Journal of Applied Business and Economics vol. 14(3) 2013
non-interest income and risk-adjusted market returns, we are interested in whether non-interest income is
related to accounting rates of return.
Table 5 displays the results of equation (3) re-estimated using accounting returns (ROA and ROE) in
place of RETURNS as the dependent variable. We find a strong statistical relationship between NII and
ROA (first panel) and also between NII and ROE (second panel). Based on our column [a] estimates, a
one-standard-deviation increase in NII is associated with a 76 percent increase in ROA (84 basis points)
and a 41 percent increase in ROE (347 basis points) at the means of the data. Indeed, the evidence here
suggests that had an accounting-based financial analysis been performed during our sample period, there
would have appeared to be a risk-adjusted return premium associated with non-interest income. This is
consistent with the conventional wisdom at the time among industry participants that expansion into noninterest activities resulted in improved risk–return trade-offs at commercial banks.
Subsamples by Time
Non-interest income has been increasing as a percentage of commercial bank income in the US since
the early 1980s (e.g., DeYoung & Rice, 2004; Kaufman & Mote, 1994,). Figure 1 shows that non-interest
income increased for both US and non-US commercial banks during our 1995–2002 sample period,
although the increase moderated in the later years. This change in the income composition has required
banks to conduct many things differently from in the past, including using new production processes, new
financial products and services, and new pricing methods. As discussed above, it is likely that these
changes affected banks’ risk profiles through changes in their revenue streams, operating and/or financial
leverage, exposure to interest rate risk, etc. It is also likely that these continuous innovations made it
difficult at first for market investors to price commercial bank equity shares accurately; risk-averse
investors with little knowledge of the true risk profiles of these new business methods may have initially
priced-in a large risk premium, and relaxed this stance after several years of financial performance data
observations made the risks associated with these new methods more transparent.11
We re-estimated our system of equations (2, 3) after splitting the data into an early time period
(1995–1998) and a later time period (1999–2001). The results are shown in Tables 6 and 7. These
subsample regressions are important tests of robustness, as they are largely consistent with our main fullsample results (see Tables 2 and 3) in both the early and the later subsamples. In the equation (2)
regressions displayed in Table 6, MKTRISK is positively and significantly associated with NII in both
time periods, while TOTRISK and IDIORISK continue to be statistically unrelated to NII in both time
periods. In the equation (3) regressions displayed in Table 7, there is a positive estimated trade-off
between RETURN and TOTRISK, MKTRISK, and IDIORISK in both time periods, and no statistical
relationships between RETURN and NII in either time period.
Moreover, these regressions are consistent with (though by no means constitute strong proof of) our
above conjecture that investors reduced their assessments of the riskiness of non-interest income as time
passed. For example, in Table 6, the positive coefficient for NII declines in size, suggesting that the
market scaled back its perception of the riskiness of non-interest income as time passed. Similarly, the
market associated non-interest income with decreased interest rate risk early in the sample period, but this
belief disappeared in the later period. The data in Table 7 tell a similar story. The positive risk–return
trade-offs declined by about 50 percent for TOTRISK and IDIORISK, and by about 25 percent for
MKTRISK, as time elapsed.12
Additional Robustness Tests
We performed a number of additional robustness tests of the main results displayed in Tables 2 and 3.
When we imposed bank fixed effects on the model, the signs and statistical significance levels for all the
main test coefficients, NII in equation (2), and NII and RISK in equation (3), were unchanged. The results
are not reported here, but they are available from the authors upon request.
Journal of Applied Business and Economics vol. 14(3) 2013
133
CROSS-COUNTRY RESULTS
To investigate further the relationship between the non-traditional activities of a bank and its risk and
return, in this section we examine the extent to which the institutional environment affects the
relationship. Many economists have argued that bank-based systems are better at mobilizing funds,
managing risk, and exerting sound corporate control, especially in weak institutional environments
(Levine, 1997). Calomiris (1994) showed that the historical evidence indicates that German universal
banks (with a relatively high level of non-traditional activities) were more efficient (lower cost of capital)
than US banks and suffered fewer systemic problems than the US banking system. On the other hand, if a
bank is operating in a market-based economy, its positive role is relatively limited and results in low
demand for non-traditional activities. Therefore, banks are not good at managing risk due to non-interest
income activities. Tables 8A and 8B show the results. In a market-based country, NII has a significant and
positive association with the market risk. We do not find any impact of NII on risk in a bank-based
country. In Table 8B, we find a more interesting result. In a market-based country, there is a significantly
positive risk–return trade-off. On the other hand, we do not find a significant risk–return trade-off in a
bank-based country. The result, however, shows that NII has a significantly positive relationship with
stock returns. This suggests that the market efficiently prices the risk–return trade-off in a market-based
economy, but not in a bank-based economy.
The risk of non-traditional activities can be contained through effective and prudential regulation and
supervision. Depositors frequently lack the incentives and capabilities to monitor and discipline banks.
Hence, governmental regulation and supervision may reduce the information asymmetries and are often
essential to ensure the solvency of the whole banking system. The stronger supervisory power and strict
regulation will improve the governance of banks by direct monitoring and discipline and will increase the
likelihood that banks will allocate resources efficiently based on risk–return trade-offs. The enhanced
governance of banks through strengthened supervisory power and stringent regulation may reduce the
moral hazard of banks engaging in non-traditional activities. Therefore, tougher supervision and
regulation lead to risk reduction. On the other hand, supervisors may maximize their private welfare
instead of maximizing more social welfare, which preserves the safety and soundness of the banking
system. In addition, tighter regulation and more supervision may reduce the competition, efficiency, and
competitiveness of the banking system. This view, therefore, suggests that the risk profile due to nontraditional activities may differ among the levels of supervision and regulation. We use the bank freedom
index as our main variable to test the potential different effects of NII on risk and return. Tables 9A and
9B document the findings. We find that NII has a significant and positive association with the total and
market risk if the bank freedom is high. In contrast, there is a negative association between NII and risk in
a low-bank-freedom country. The results, therefore, suggest that stronger supervisory power and strict
regulation will improve governance and lead to risk reduction. In Table 9B, we find a significant risk–
return trade-off in both countries.
The deposit insurance, especially explicit deposit insurance (EDI), reduces the losses that depositors
incur in the case of bank failure. However, having an explicit deposit insurance scheme may lead to
greater moral hazard for bank managers, who may take advantage of the deposit insurance program by
engaging in more activities that may increase risk. The banking literature suggests that the more generous
deposit insurance is, the greater are the risk-taking incentives for banks. Deposit insurance may make
depositors less likely to enforce market discipline on banks and may induce banks to take additional
risks.13 The empirical evidence also shows that deposit insurance increases the probability of banking
crisis (Demirguc-Kunt & Detragiache, 2002; Demirguc-Kunt & Huizinga, 2004). A more generous
deposit insurance scheme may therefore lead to greater moral hazard for the bank managers. Bank
managers may take advantage of the explicit deposit insurance by greatly engaging in non-traditional
activities, which may be riskier than traditional lending activities. Tables 10A and 10B show the results
based on the existence of EDI. Consistent with the literature, we find a strong and positive association
between NII and market risk in a country with EDI, while we find a negative relationship in a country
134
Journal of Applied Business and Economics vol. 14(3) 2013
without EDI. The finding suggests that explicit deposit insurance increases the moral hazard and thereby
induces the risk-taking incentives for banks, resulting in a great risk to banks.
Finally, concentration (CONCEN) is the share of assets of the three largest banks in a country. If the
banking sector is concentrated due to regulation, few large banks may enjoy rents. Bank concentration
may lead to a “too-big-to-fail” policy and depositors do not have to care about bank failure. Bank
concentration may also lead banks to engage in more non-traditional activities by exploiting the implicit
guarantee from the government. We include a measure of banking concentration of each country from the
World Bank Database (2004). This is the measure of the fraction of assets in the five largest banks that is
owned by commercial banks and/or financial conglomerates. If there are fewer than five banks, it uses
that number to calculate the index. Bank concentration may lead to the TBTF policy and depositors not
having to care about bank failure. Bank concentration may also lead banks to engage in more risky
activities by exploiting the implicit guarantee from the government. Tables 11A and 11B show the results
based on the concentration. We find a strong and positive association between NII and market risk in a
less concentrated market. The finding suggests that concentration leads banks to engage in more nontraditional activities by exploiting the implicit guarantee from the government and it results in a great risk
to banks.
CONCLUSION
Even though there is evidence of decreasing traditional activities among banks, the empirical studies
on banks have mainly analyzed the role of the banks in terms of the traditional activities generating
interest income. Kaufman and Mote (1994) argued that the nature of banking activities changed steadily
in the 1990s. DeYoung and Roland (2001) showed that non-interest income at FDIC-insured commercial
banks increased from 25% to over 40% of their aggregate income over the period 1984 to 2001. This is
not only a trend for US banks, but also a trend for banks worldwide. The growth of non-interest income
seems to have a positive effect on bank profitability and few studies have found consistent evidence on
this trend in European countries (Smith et al., 2003).
Although there is a general belief that fee-based earnings are more stable than loan-based income,
DeYoung and Roland (2001) provided three potential observations that explain that fee-based income
may not be more stable than income from traditional banking activities: the low switching and
information costs, the need to hire an additional fixed labor input (more expense ratio), and the no
regulatory capital requirement that banks do not need to hold capital against fee-based activities (higher
degree of leverage). The empirical results of the effects of non-traditional banking activities on the
riskiness of bank are mixed. It is therefore interesting and important to know whether and how these new
activities affect the nature of riskiness and the performance of banks.
This paper investigates the effect of non-interest income on the riskiness and profitability of banks.
Specifically, using the data of 877 banks in 42 countries, we investigate whether non-traditional activities
of banking institutions affect the risk and returns of banks. This is the first paper to examine this issue in a
global context using capital market measures of risk as well as market and book value measures for
performance. In general, we find that non-interest income is associated with riskier stock returns at
commercial banking companies, due primarily to increased market, or systematic, risk. These findings are
new to the literature, and they suggest that fee-based banking increases banks’ exposure to the business
cycle. In contrast, we find almost no evidence linking non-interest income to changes in other risk
measures, such as total risk, interest rate risk, or idiosyncratic risk. In addition, our findings suggest that,
on average, the stock markets efficiently price the increased risk associated with the non-interest income
market. That is, after controlling for cross-sectional differences in risk, market returns do not fluctuate
with the mix of bank income. However, we find evidence of a premium in accounting returns that is
positively related to non-interest income. This set of results is also new to the literature, and offers a
potential explanation for the initial conventional wisdom among industry participants that expansion into
non-interest activities would result in an improved risk–return trade-off at commercial banks. Finally, we
find that cross-country differences in regulatory practices, economic conditions, and social institutions
Journal of Applied Business and Economics vol. 14(3) 2013
135
influence our main results in important ways, but on average our risk–return results appear to be robust
across countries.
Our findings have several policy implications for regulation. Non-traditional activities can be less
transparent than traditional banking activities, thus they are more difficult to monitor for safety and
soundness. For example, reductions in equity capital caused by credit losses are obvious and easily
identifiable (i.e., the progression of loan delinquency, loan classification, loan provisioning, and loan
charge-off), but shocks to non-interest income affect equity capital more subtly through ex post
reductions in retained earnings, which for most established firms are the primary source of capital.
Activities that generate volatile earnings streams make this source of capital riskier, and this volatility is
exacerbated by financial leverage. Under the current regulatory capital rules, banks are not required to
hold capital against most fee-generating activities. Our findings invite a discussion about whether the risk
associated with certain non-interest activities is large enough and systematic enough to merit a required
capital charge in future versions of these capital regulations. Our findings also have implications for
investors. If the additional risk associated with non-interest income is predominantly systematic risk (as
we find here), then investors will require higher expected returns to accept this non-diversifiable risk in
their portfolios.
ENDNOTES
1.
2.
3.
4.
5.
6.
7.
8.
9.
136
Based on Federal Reserve Y-9C filings. We define net revenue as net interest income plus non-interest
income. See Table 1 and Figure 1 below for some trends in non-interest income for non-US banking
companies. Early studies documenting increases in non-interest income at commercial banks include those
by Boyd and Gertler (1994) and Kaufman and Mote (1994). Choi (2005) documents the differences in the
growth and determinants of non-interest income across banks in 42 different nations.
This organizational dichotomy follows the analysis of DeYoung and Rice (2004, Table 1).
In the US, the barriers between banking, securities, and insurance activities were reduced on an ad hoc
basis during the late 1980s and early 1990s; the crowning blow was the Gramm–Leach–Bliley (GLB) Act
of 1999, which allowed banking and non-banking activities to be affiliated within a single financial holding
company. In the EU, the Second Banking Co-ordination Directive of 1989 (implemented in 1993–94)
permitted universal banks to expand anywhere within the EU regardless of the host country restrictions on
product powers.
The elimination of deposit interest-rate ceilings in many countries has also led to increased fee income
from depositor services, by allowing banks to price depositor services in a more rational and competitive
fashion.
The yet-to-be-implemented Basel II capital framework includes capital charges on various fee-generating
lines of business, but it uses largely ad hoc risk weights with no rigorous justification for the magnitudes.
There is also a capital charge for operational risk, which is determined by the level of total operating (net
interest plus non-interest) income.
Equation (2) is estimated by itself, and then equation (2) is estimated using the fitted value of RISK from
(1) as a right-hand-side instrumental variable.
DeLong and DeYoung (2007) provided evidence that stock investors learn with experience how better to
price new phenomena such as large, complex bank M&As.
We excluded central banks, cooperative banks, investment banks, Islamic banks, medium- and long-term
credit banks, non-banking credit institutions, real estate/mortgage banks, and specialized governmental
credit institutions.
The KKZ index is the aggregate indicators of six dimensions of governance: 1. Voice and Accountability—
measuring political, civil, and human rights; 2. Political Instability and Violence—measuring the likelihood
of violent threats to, or changes in, government, including terrorism; 3. Government Effectiveness—
measuring the competence of the bureaucracy and the quality of public service delivery; 4. Regulatory
Burden—measuring the incidence of market-unfriendly policies; 5. Rule of Law—measuring the quality of
contract enforcement, the police, and the courts, as well as the likelihood of crime and violence; 6. Control
of Corruption—measuring the exercise of public power for private gain, including both petty and grand
corruption and state capture. The indicators are constructed using an unobserved components methodology
described in detail in the paper. The index is measured ranging from about -2.5 to 2.5, with higher values
Journal of Applied Business and Economics vol. 14(3) 2013
10.
11.
12.
13.
corresponding to better governance. References: Kaufmann, D., Kraay, A., & Zoido-Lobaton, P. (1999a).
Aggregating Governance Indicators. World Bank Policy Research Department Working Paper No. 2195;
Kaufmann, D., Kraay, A., & Zoido-Lobaton, P. (1999a). Governance Matters. World Bank Policy
Research Department Working Paper No. 2196.
The index is from the Heritage Foundation and the WSJ—the Heritage Foundation/Wall Street Journal
Index of Economic Freedom.
Recent studies by DeLong and DeYoung (2007) and Pastor and Veronesi (2003) provided empirical
evidence linking stock market prices to investor learning.
We acknowledge that other phenomena may be partially or wholly responsible for the observed reductions
in the perception and pricing of risk associated with non-interest income in Tables 6 and 7. For example,
the mix of financial services that generated the non-interest income in the two subsample periods may have
been different.
See Bhattacharya, Boot, and Thakor (1998) and Demirguc-Kunt and Detragiache (2002).
REFERENCES
Allen, L.G., & Jagtiani, J. (2000). The Risk Effects of Combining Banking, Securities, and Insurance
Activities. Journal of Economics and Business, 52, 485–497.
Barth, J.R., Caprio Jr., G., & Levine, R. (2001). The Regulation and Supervision of Bank around World:
A New Fatabase. In R.E. Litan, & R. Herring (Eds.), Integrating Emerging Market Countries into the
Global Financial System. Washington, DC: Brooking Institution Press.
Bhattacharya, S., Boot, A., & Thakor, A.V. (1998). The Economics of Bank Regulation. Journal of
Money, Credit & Banking, 30, 745–770.
Boyd, J.H., & Gertler, M. (1994). Are banks dead? Or are the reports greatly exaggerated? Quarterly
Review, Federal Reserve Bank of Minneapolis, 2-23.
Boyd, J. H., Hanweck, G. A., and Pithyachariyakul, P. (1980). Bank Holding Company Diversification,
Federal Reserve Bank of Chicago, Proceedings from a Conference on Bank Structure and Competition,
105-120.
Boyd, J.H., & Graham, S.L. (1986). Risk, Regulation, and Bank Holding Company Expansion into
Nonbanking. Federal Reserve Bank of Minneapolis Quarterly Review, 10 (2), 2–17.
Boyd, J.H., Graham, S.L., & Hewitt, R.S. (1993). Bank Holding Company Mergers with Nonbank
Financial Firms: Effects on the Risk of Failure. Journal of Banking & Finance, 17, 43–63.
Brewer III, E. (1989). Relationship between Bank Holding Company Risk and Nonbank Activity. Journal
of Economics and Business, 41 (4), 337–353.
Calomiris, C.W. (1994). The Costs of Rejecting Universal Banking: American Finance in German Mirror,
1870–1914. In N. Lamoreau, & D. Raff (Eds.), The Coordination of Economic Activities Within and
Between Firms. Chicago: University of Chicago Press.
Choi, S. (2005). Three Essays in Banking and Finance. Rensselaer Polytechnic Institute, PhD
Dissertation.
Cooper, M.J., Jackson III, W.E., & Patterson, G.A. (2003). Evidence of Predictability in the CrossSection of Bank Stock Returns. Journal of Banking & Finance, 27, 817–850.
Journal of Applied Business and Economics vol. 14(3) 2013
137
DeLong, G. and DeYoung, R. (2007), Learning by Observing: Information Spillovers in the Execution
and Valuation of Commercial Bank M&As. The Journal of Finance, 62, 181–216. doi: 10.1111/j.15406261.2007.01205.x
DeYoung, R., Hunter, W., & Udell, G. (2004). The Past, Present, and Probable Future for Community
Banks. Journal of Financial Services Research, 25(2), 85-133.
DeYoung, R., & Roland, K.P. (2001). Product Mix and Earnings Volatility at Commercial Banks:
Evidence from a Degree of Leverage Model. Journal of Financial Intermediation, 10, 54–84.
DeYoung, R., & Rice T. (2004). Noninterest Income and Financial Performance at U.S. Commercial
Banks. Financial Review, 39, 101-127. doi: 10.1111/j.0732-8516.2004.00069.x
Demirguc-Kunt, A., & Detragiache, E. (2002). Does Deposit Insurance Increase Banking System
Stability? An Empirical Investigation. Journal of Monetary Economics, 49, 1373–1406.
Demirguc-Kunt, A., & Huizinga, H. (2004). Market Discipline and Deposit Insurance. Journal of
Monetary Economics, 51, 375–399.
Demsetz, R.S., & Strahan, P.E. (1997). Diversification, Size, and Risk at Bank Holding Companies.
Journal of Money, Credit, and Banking, 29, 300–313.
Eisenbeis, R.A., Harris, R.S., & Lakonishok, J. (1984). Benefits of Bank Diversification: The Evidence
from Shareholder Returns. Journal of Finance, 34, 881–892.
Esho, N., Kofman, P., & Sharpe., I. G. (2005). Diversification, Fee Income, and Credit Union Risk.
Journal of Financial Services Research, 27(3), 259-281.
Gallo, J.G., Apilado, V.P., & Kolari, J.W. (1996). Commercial Bank Mutual Fund Activities:
Implications for Bank Risk and Profitability. Journal of Banking & Finance, 20, 1775–1791.
Heggestad, A. (1975). Riskiness of Investments in Nonbank Activities by Bank Holding Companies.
Journal of Economics & Business, 219–223.
Johnson R.N. & Meinster, D.R. (1974). Bank Holding Companies: Diversification Opportunities in
Nonbank Activities. Eastern Economic Journal, 1, 316-323.
Kaufmann, D., Kraay, A., & Zoido-Lobaton, P. (1999). Aggregating Governance Indicators. Manuscript,
The World Bank.
Kaufman, G.G., & Mote, L.R. (1994). Is Banking a Declining Industry? A Historical Perspective. FBR
NY Economic Perspectives, 18, 2–21.
Kwan, S.H. (1997). Securities Activities by Commercial Banking Firms’ Section 20 Subsidiaries: Risk,
Return, and Diversification Benefits. Working paper, Federal Reserve Bank of San Francisco.
Kwast, M.L. (1989). The Impact of Underwriting and Dealing on Bank Returns and Risks. Journal of
Banking and Finance, 13, 101–125.
Laderman, E. S. (1998). The Potential Diversification and Failure Reduction Benefits of Bank Expansion
into Nonbanking Activities. Federal Reserve Bank of San Francisco, Working Paper 2000-01.
138
Journal of Applied Business and Economics vol. 14(3) 2013
Levine, R. (1997). Financial Development and Economic Growth: Views and Agenda. Journal of
Economic Literature, 688–726.
Litan, R.E. (1985). Evaluating and Controlling the Risks of Financial Product Deregulation. Yale Journal
of Regulation, 3, 1-52.
Pastor, L., & Veronesi, P.. (2003). Stock Valuation and Learning about Profitability. Journal of Finance,
58:1749–89.
Rogers, K.E., & Sinkey, J.F., Jr. (1999). An Analysis of Nontraditional Activities at US Commercial
Banks. Review of Financial Economics, 8, 25–39.
Roland, K. P. (1997). Profit Persistence in Large U.S. Bank Holding Companies: An Empirical
Investigation. Office of the Comptroller of the Currency, Economics Working Paper, 97-2.
Rosen, R. J., Lloyd-Davies, P. R., Kwast, M.L., & Humphrey, D. B. (1989). New Banking Powers.
Journal of Banking and Finance, 13, 355–366.
Sinkey J., & Nash, R. (1993). Assessing the Riskiness and Profitability of Credit Card Banks. Journal of
Financial Services Research, 7, 127-150.
Smith, R., Staikouras, C., & Wood, G. (2003). Noninterest Income and Total Income Stability. Working
paper, Bank of England.
Stiroh, K.J. (2004a). Diversification in Banking: Is Noninterest Income the Answer? Journal of Money,
Credit, and Banking, 36(5), 853-882.
Stiroh, K.J. (2004b). Do Community Banks Benefit from Diversification? Journal of Financial Services
Research, 25(2-3), 135-160. doi: 10.1023/B:FINA.0000020657.59334.76
Wall, L.D. & Eisenbeis, R. A.(1984). Risk Considerations in Deregulating Bank Activities. Federal
Reserve Bank of Atlanta, Economic Review, 69, 6-18.
Wall, L. D., Reichert, A. K., & Mohanty, S. (1993). Deregulation and the Opportunities for Commercial
Bank Diversification. Federal Reserve Bank Atlanta Economic Review, 1-25.
Journal of Applied Business and Economics vol. 14(3) 2013
139
TABLE 1
SUMMARY STATISTICS
Variable
Definition
Mean
Median
Standard
Deviation
Minimum
Maximum
Variables used to estimate the market model, equation (1)
Return of an individual bank
-0.0001
0.0000
0.0262
-0.3966
0.5576
MR
Market return of a country
-0.0002
0.0000
0.0175
-0.1941
0.2246
INT
Change in interest rate in a
country
-0.0000
0.0000
0.0014
-0.0162
0.0171
R
RISK, RETURN, and NII variables used in equations (2) and (3)
TOTRISK
Total risk estimated from market
model (1)
0.0228
0.0205
0.0145
0.0008
0.3945
MKTRISK
Market risk estimated from
market model (1)
0.4533
0.3544
0.4624
-1.3422
3.0279
IRRISK
Interest rate risk estimated from
market model (1)
-0.3268
-0.1892
3.4478
-17.7642
28.5523
IDIORISK
Idiosyncratic risk estimated from
market model (1)
0.0207
0.0186
0.0119
0.0008
0.1730
RETURN
Annual stock return (annualized
average of weekly returns)
0.0346
0.0232
0.1089
-0.3732
2.0467
ROA
Return on book assets
0.0111
0.0101
0.0267
-0.4820
0.5608
ROE
Return on book equity
0.0849
0.1049
0.1752
-3.5711
1.2082
-0.0627
0.3926
-0.0738
1.0000
6,887
1,097,190,000
8.8374
0.0002
20.8160
0.9704
0.0002
0.9185
0.0000
0.1811
0.0000
1.0000
0.2000
1.0000
0.0012
0.9071
NII
NII(alt)
ASSETS
lnASSETS
LOANS
LIQUID
LOSSPROV
EXPLICIT
CR3
CAPITAL
140
Non-interest income divided by
0.0185
0.0116
0.0307
total assets
Non-interest income divided by
operating revenue (non-interest
0.1901
0.1538
0.1568
income plus net interest income)
Control variables used in equations (2) and (3)
Total assets, expressed in
22,315,504
2,684,594
71,483,148
thousands of US dollars
Natural log of ASSETS
14.9887
14.8030
1.9493
Total loans divided by total assets
0.6016
0.6297
0.1618
Liquid assets divided by total
0.1605
0.0988
0.1494
assets
Loan loss provisions divided by
0.0053
0.0029
0.0094
total assets
Dummy = 1 for explicit deposit
0.8874
1.0000
0.3162
insurance
Three-bank concentration ratio
0.3652
0.2700
0.2060
Market value of bank equity
0.1038
0.0822
0.1177
divided by book value of assets
Journal of Applied Business and Economics vol. 14(3) 2013
RESTRICT
CORPGOV
BANKFREE
MULTSUPS
MARKET
FOREIGN
STATE
Increasing index of permissible
bank activities. Barth et al. (2001)
Increasing index of corporate
governance. Kaufman, Kraay,
and Zoido-Lobaton (KKZ)
Index of bank freedom of the
country
Dummy = 1 if multiple banking
supervisors
Dummy = 1 if economy is
market-based (versus bank-based)
10.3558
12.0000
2.5738
3.0000
14.0000
1.1070
1.2900
0.4514
-0.7600
1.7200
3.7216
4.0000
0.5655
2.0000
5.0000
0.5914
1.0000
0.4916
0.0000
1.0000
0.6858
1.0000
0.4642
0.0000
1.0000
Dummy = 1 if bank is a foreignowned bank
0.0416
0.0000
0.1998
0.0000
1.0000
Dummy = 1 if bank is a stateowned bank
0.0209
0.0000
0.1432
0.0000
1.0000
The Fitch-IBCA Bankscope database is our primary source for annual bank-level financial data. The bank
stock returns Rit, stock market indices MRit, and interest rates INTit come from Datastream. Appendix I
contains additional details about the data sources for MR. Datastream does not report long-term
government bond yields for all countries; in these cases, we use a long-term corporate bond yield to
calculate INT. The remainder of the control variables are observed from various sources.
Journal of Applied Business and Economics vol. 14(3) 2013
141
TABLE 2
MAIN REGRESSIONS, EQUATION (2)
RISKit = μ + λNIIit + ψYit + τΣt=1,TTt + πit
Dependent
variable:
Constant
NII
LOANS
lnASSETS
CAPITAL
LIQUID
LOSSPROV
MARKET
EXPLICIT
RESTRICT
CORPGOV
BNKFREE
CR3
MULTSUPS
YR95
YR96
YR97
YR98
YR99
YR00
YR01
Adj-R2
N
142
[a]
[b]
[c]
TOTRISK
coefficient
p-value
0.0424
(.000)
0.0058
(.396)
-0.0073
(.001)
-0.0015
(.000)
0.0023
(.002)
0.0061
(.021)
0.1624
(.000)
0.0002
(.829)
-0.0019
(.138)
0.0005
(.000)
-0.0077
(.000)
-0.0020
(.000)
-0.0019
(.195)
0.0061
(.000)
-0.0014
(.233)
-0.0030
(.000)
0.0009
(.159)
0.0048
(.000)
0.0031
(.000)
0.0048
(.000)
0.0005
(.320)
0.2313
4444
MKTRISK
coefficient
p-value
-1.205
(.000)
1.0072
(.000)
-0.1785
(.000)
0.3008
(.000)
0.5847
(.000)
0.1501
(.009)
1.3610
(.166)
0.0638
(.004)
-0.1258
(.000)
0.0158
(.000)
-0.2173
(.000)
-0.1097
(.000)
0.4446
(.000)
0.1747
(.000)
-0.0810
(.003)
-0.0257
(.287)
0.0297
(.156)
0.1056
(.000)
-0.0547
(.003)
-0.0890
(.000)
-0.0267
(.145)
0.4728
4444
IRRISK
coefficient
p-value
6.4148
(.000)
0.4166
(.873)
-1.3412
(.004)
-0.4921
(.000)
-0.0442
(.936)
-0.3906
(.504)
6.4626
(.388)
-0.0684
(.660)
-0.0360
(.866)
-0.1626
(.000)
-0.1428
(.388)
-0.0227
(.853)
-0.8500
(.061)
0.1291
(.439)
-1.2639
(.000)
-1.1179
(.000)
-1.2797
(.000)
-0.5676
(.004)
-0.7478
(.000)
-0.5266
(.014)
-0.1371
(.441)
0.0380
4444
Journal of Applied Business and Economics vol. 14(3) 2013
[d]
IDIORISK
coefficient
p-value
0.0476
(.000)
0.0053
(.400)
-0.0052
(.000)
-0.0029
(.000)
-0.0012
(.430)
0.0034
(.056)
0.1332
(.000)
0.0002
(.781)
-0.0006
(.262)
0.0004
(.000)
-0.0068
(.000)
-0.0010
(.028)
-0.0075
(.000)
0.0037
(.000)
-0.0009
(.133)
-0.0015
(.005)
0.0014
(.016)
0.0044
(.000)
0.0040
(.000)
0.0060
(.000)
0.0012
(.016)
0.2541
4444
TABLE 3
MAIN REGRESSIONS, EQUATION (3)
RETURNit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
Dependent
variable:
Constant
NII
LOANS
lnASSETS
TOTRISK
MKTRISK
IRRISK
IDIORISK
LIQUID
LOSSPROV
EXPLICIT
FOREIGN
STATE
CR3
YR95
YR96
YR97
YR98
YR99
YR00
YR01
Adj-R
N
(a)
(b)
(c)
RETURN
coefficient
p-value
-0.0665
(.168)
0.2486
(.145)
-0.0356
(.098)
0.0010
(.761)
4.6876
(.000)
RETURN
coefficient
p-value
0.3061
(.000)
0.0296
(.869)
-0.0099
(.642)
-0.0550
(.000)
RETURN
coefficient
p-value
0.2194
(.000)
0.2845
(.107)
-0.0942
(.000)
-0.0154
(.002)
0.1770
-0.0993
-2.1926
-0.0037
-0.0290
-0.0672
0.1129
0.0170
0.0361
0.0525
-0.0205
-0.0059
-0.0332
0.0063
0.0606
4444
(.002)
(.000)
(.749)
(.032)
(.000)
(.000)
(.001)
(.000)
(.000)
(.009)
(.382)
(.000)
(.326)
-0.0735
-1.6851
0.0180
-0.0308
-0.0734
0.0290
0.0253
0.0274
0.0515
-0.0162
0.0186
0.0051
0.0133
0.0715
4444
(d)
RETURN
coefficient
p-value
-0.1090
(.076)
0.2622
(.125)
-0.0456
(.038)
0.0090
(.065)
(.000)
(.013)
(.000)
(.111)
(.017)
(.000)
(.016)
(.000)
(.000)
(.000)
(.024)
(.027)
(.402)
(.057)
-0.0151
(.206)
-0.0374
-0.9360
-0.0382
-0.0091
-0.0475
0.0327
-0.0102
0.0044
0.0383
-0.0067
-0.0028
-0.0183
0.0068
0.0460
4444
(.138)
(.001)
(.008)
(.466)
(.000)
(.011)
(.262)
(.613)
(.005)
(.413)
(.728)
(.000)
(.000)
5.3747
-0.0914
-2.1386
-0.0132
-0.0256
-0.0647
0.1291
0.0146
0.0295
0.0488
-0.0221
-0.0136
-0.0431
0.0020
0.0581
4444
Journal of Applied Business and Economics vol. 14(3) 2013
(.000)
(.004)
(.000)
(.279)
(.057)
(.000)
(.000)
(.005)
(.000)
(.000)
(.008)
(.068)
(.000)
(.750)
143
TABLE 4
ALTERNATE SPECIFICATION OF EQUATION (3), SELECTED RESULTS
RETURNit = ν + φNIIit + δRISKit + φNIIit*RISKit + ρZit + τΣt=1,TTt + ηit
Dependent variable:
NII
TOTRISK
TOTRISK*NII
MKTRISK
MKTRISK*NII
IRRISK
IRRISK*NII
IDIORISK*
IDIORISK*NII
(a)
(b)
(c)
RETURN
coefficient
p-value
-0.2036
(.052)
2.3354
(.019)
10.8248
(.473)
RETURN
coefficient
p-value
-0.8503
(.009)
RETURN
coefficient
p-value
0.3248
(.046)
0.1252
1.1036
0.0432
Adj-R
N
0.0481
4444
144
[2.07]
-0.3500
(.770)
2.5356
[2.88]
[-3.04]
0.0880
(.005)
0.1456
(.005)
RETURN
coefficient
p-value
-0.0215
(.943)
(.004)
(.075)
-0.0152
-0.1699
∂RETURN/∂NII
Joint significance of
NII and RISK*NII
∂RETURN/∂RISK
Joint significance of
RISK and RISK*NII
(d)
[3.55]
Journal of Applied Business and Economics vol. 14(3) 2013
[2.26]
2.7180
5.4947
(.018)
(.722)
0.0923
[0.74]
(.030)
-0.0184
(.002)
0.0686
4444
(.025)
(.047)
[-2.84]
(.7558)
2.8196
(.002)
0.0444
4444
[2.71]
(.014)
0.0461
4444
TABLE 5
ALTERNATE SPECIFICATION OF EQUATION (3), SELECTED RESULTS
ROAit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
Dependent
variable:
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R
N
(a)
(b)
(c)
ROA
coefficient
p-value
0.3120
(.000)
0.1282
(.413)
ROA
coefficient
p-value
0.2756
(.000)
ROA
coefficient
p-value
0.3095
(.000)
0.0282
(d)
ROA
coefficient
p-value
0.3123
(.000)
(.000)
0.0017
(.209)
-0.0964
0.2348
4444
0.2566
4444
0.2350
4444
(.643)
0.2219
4375
ROEit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
Dependent
variable:
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R
N
(a)
(b)
(c)
ROE
coefficient
p-value
1.1119
(.000)
0.9363
(.401)
ROE
coefficient
p-value
1.0512
(.000)
ROE
coefficient
p-value
1.1045
(.000)
0.0484
(d)
ROE
coefficient
p-value
1.1144
(.000)
(.086)
0.0058
(.464)
0.6479
0.2601
4444
0.2160
4444
0.2598
4444
(.648)
0.2599
4444
Journal of Applied Business and Economics vol. 14(3) 2013
145
TABLE 6
EQUATION (2) FOR THE 1995–1998 AND 1999–2002 SUBSAMPLES
RISKit = μ + λNIIit + ψYit + τΣt=1,TTt + πit
Dependent
variable:
(a)
(b)
(c)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
NII
Adj-R2
N
-0.0093
0.2537
1791
(0.331)
1.0869
0.4612
1791
1995–1998
(0.022)
-1.4021
0.0444
1791
NII
Adj-R2
N
0.0124
0.2293
2653
(0.177)
0.9556
0.4998
2653
1999–2002
(0.001)
1.4387
0.0307
2653
(d)
IDIORISK
coefficient
p-value
(0.595)
-0.0038
0.3410
1791
(0.595)
(0.701)
0.0092
0.2135
2653
(0.296)
TABLE 7
EQUATION (3) FOR THE 1995–1998 AND 1999–2002 SUBSAMPLES
RETURNit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
Dependent
variable:
(a)
(b)
(c)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
0.1902
2.6354
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
0.2558
5.7614
146
(0.247)
(0.134)
0.0210
1995–1998
(0.911)
0.1525
0.0886
(0.128)
0.0138
0.0431
1791
0.0459
1791
(0.289)
(0.000)
0.0553
0.0386
1791
1999–2002
(0.828)
0.4233
0.2169
(0.000)
-0.0382
0.0839
2653
0.0988
2653
Journal of Applied Business and Economics vol. 14(3) 2013
0.0718
2653
(d)
(0.356)
IDIORISK
coefficient
p-value
0.1850
(0.262)
2.6011
0.0653
1443
(0.206)
0.2781
(0.249)
7.2757
0.0812
2653
(0.000)
(0.324)
(0.094)
(0.000)
TABLE 8A
EQUATION (2) FOR THE MARKET VS BANK-BASED SUBSAMPLES
RISKit = μ + λNIIit + ψYit + τΣt=1,TTt + πit
Dependent variable:
(a)
(b)
(c)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
NII
Adj-R2
N
-0.0027
0.2676
3057
(0.711)
0.7813
0.5490
3057
NII
Adj-R2
N
-0.0055
0.2292
1387
(0.769)
0.9920
0.4460
1387
Market
(0.001)
0.9233
0.0537
3057
(0.111)
Bank
3.3694
0.0410
1387
(d)
IDIORISK
coefficient
p-value
(0.792)
-0.0025
0.2345
3057
(0.710)
(0.345)
0.0041
0.3374
1387
(0.754)
TABLE 8B
EQUATION (3) FOR THE MARKET VS BANK-BASED SUBSAMPLES
RETURNit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
Dependent
variable:
(a)
(b)
(c)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
0.2974
9.4899
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
0.2599
0.3342
(0.228)
(0.000)
0.0463
Market
(0.863)
0.3763
0.2593
(0.001)
-0.0330
0.1047
3057
0.0875
3057
(0.077)
(0.656)
0.2499
0.0656
3057
Bank
(0.089)
0.3249
0.0007
(0.969)
-0.0119
0.1322
1387
0.1320
1387
0.1340
1387
(d)
(0.140)
IDIORISK
coefficient
p-value
0.3133
(0.203)
13.0544
0.1048
3057
(0.000)
0.2628
(0.073)
0.4761
0.1323
1387
(0.570)
(0.000)
(0.032)
(0.067)
Journal of Applied Business and Economics vol. 14(3) 2013
147
TABLE 9A
EQUATION (2) FOR THE BANK FREEDOM SUBSAMPLES
RISKit = μ + λNIIit + ψYit + τΣt=1,TTt + πit
Dependent
variable:
(a)
(b)
(c)
(d)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
IDIORISK
coefficient
p-value
High bank freedom
NII
Adj-R2
N
0.0180
0.2884
2995
(0.059)
1.7571
0.5560
2995
NII
Adj-R2
N
-0.0360
0.1981
1449
(0.002)
-0.4942
0.3952
1449
(0.000)
1.4354
0.0649
2995
(0.704)
0.0101
0.2727
2995
(0.273)
(0.200)
-0.0204
0.2659
1449
(0.016)
Low bank freedom
(0.163)
3.5523
0.0348
1449
TABLE 9B
EQUATION (3) FOR THE BANK FREEDOM SUBSAMPLES
RETURNit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
Dependent
variable:
(a)
(b)
(c)
(d)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
IDIORISK
coefficient
p-value
High bank freedom
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
-0.0182
3.9950
(0.844)
(0.012)
0.1321
2995
-0.0500
(0.654)
0.0606
(0.102)
0.1191
2995
0.0927
(0.290)
-0.0142
(0.084)
0.1147
2995
-0.0010
(0.992)
5.8796
0.1323
2995
(0.011)
0.6953
(0.094)
4.5154
0.1406
1449
(0.013)
Low bank freedom
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
148
0.7394
4.3039
0.1406
1449
(0.072)
(0.005)
0.5809
(0.138)
0.1342
(0.046)
0.1465
1449
Journal of Applied Business and Economics vol. 14(3) 2013
0.1875
(0.686)
0.0567
(0.013)
0.1452
1449
TABLE 10A
EQUATION (2) FOR THE EDI SUBSAMPLES
RISKit = μ + λNIIit + ψYit + τΣt=1,TTt + πit
Dependent variable:
(a)
(b)
(c)
(d)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
IDIORISK
coefficient
p-value
With EDI
NII
Adj-R2
N
0.0111
0.2864
3947
(0.126)
1.3824
0.5149
3947
(0.000)
NII
Adj-R2
N
-0.0687
0.1700
497
(0.002)
-3.0716
0.4646
497
(0.000)
0.2991
0.0405
3947
(0.916)
0.0047
0.2672
3947
(0.505)
(0.455)
-0.0292
0.3793
497
(0.016)
No EDI
4.1214
0.1040
497
TABLE 10B
EQUATION (3) FOR THE EDI SUBSAMPLES
RETURNit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
Dependent
variable:
(a)
(b)
(c)
(d)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
IDIORISK
coefficient
p-value
With EDI
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
0.0684
2.2981
(0.351)
(0.012)
0.0876
3947
-0.0435
(0.629)
0.0708
(0.007)
0.0856
3947
0.0765
(0.310)
0.0031
(0.733)
0.0785
3947
0.0889
(0.221)
2.6314
0.0861
3947
(0.020)
3.1363
(0.001)
34.1633
0.2393
497
(0.001)
No EDI
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
3.5572
19.0767
0.2393
497
(0.001)
(0.000)
2.6461
(0.007)
0.01274
(0.004)
0.2183
497
2.0042
(0.041)
0.0618
(0.027)
0.2173
497
Journal of Applied Business and Economics vol. 14(3) 2013
149
TABLE 11A
EQUATION (2) FOR THE CONCENTRATION SUBSAMPLES
RISKit = μ + λNIIit + ψYit + τΣt=1,TTt + πit
Dependent
variable:
(a)
(b)
(c)
(d)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
IDIORISK
coefficient
p-value
Highly concentrated
NII
Adj-R2
N
0.0086
0.3439
1602
(0.342)
0.1718
0.5044
1602
(0.628)
NII
Adj-R2
N
0.0039
0.1284
2842
(0.729)
1.6594
0.5268
2842
(0.000)
5.1497
0.0456
1602
(0.231)
0.0119
0.4175
1602
(0.151)
(0.468)
-0.0039
0.1633
2842
(0.708)
Less concentrated
-2.5819
0.0550
2842
TABLE 11B
EQUATION (3) FOR THE CONCENTRATION SUBSAMPLES
RETURNit = ν + φNIIit + δRISKit + ρZit + τΣt=1,TTt + ηit
Dependent
variable:
(a)
(b)
(c)
(d)
TOTRISK
coefficient
p-value
MKTRISK
coefficient
p-value
IRRISK
coefficient
p-value
IDIORISK
coefficient
p-value
Highly concentrated
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
0.2598
4.1318
(0.424)
(0.001)
0.0746
1602
0.2159
(0.503)
0.1635
(0.000)
0.0885
1602
-0.3570
(0.473)
0.1182
(0.035)
0.0711
1602
0.2591
(0.424)
4.3674
0.0720
1602
(0.002)
0.1746
(0.042)
5.7615
0.1748
2842
(0.038)
Less concentrated
NII
TOTRISK
MKTRISK
IRRISK
IDIORISK
Adj-R2
N
150
0.1465
1.8536
0.1736
2842
(0.086)
(0.379)
0.2158
(0.033)
-0.0205
(0.500)
0.1734
2842
Journal of Applied Business and Economics vol. 14(3) 2013
0.1669
(0.054)
0.0007
(0.873)
0.1730
2842
FIGURE 1
Noninterest Income as Percentage of Total Revenues
28%
26%
24%
ALL
22%
US
20%
NON-US
18%
16%
14%
12%
10%
1995
1996
1997
1998 1999
2000
2001
2002
Noninterest Income as Percentage of Total Assets
2.60%
2.40%
2.20%
ALL
2.00%
US
1.80%
NON-US
1.60%
1.40%
1.20%
1.00%
1995
1996
1997
1998
1999
2000
2001
2002
Noninterest Income as Percentage of Net Operating Revenues
38%
36%
34%
ALL
32%
US
30%
NON-US
28%
26%
24%
22%
20%
1995
1996
1997
1998
1999
2000
2001
2002
Journal of Applied Business and Economics vol. 14(3) 2013
151
APPENDIX I
Country
Argentina
Austria
Australia
Belgium
Brazil
Canada
Chile
Czech Republic
Denmark
Estonia
Finland
France
Germany
Greece
Hong Kong
Hungary
India
Indonesia
Ireland
Israel
Italy
Japan
Luxembourg
Malaysia
Mexico
Netherlands
Norway
Pakistan
Poland
Portugal
Singapore
South Korea
South Africa
Spain
Sri Lanka
Sweden
Switzerland
Taiwan
Thailand
Turkey
UK
US
152
Country
Code
AR
AT
AU
BE
BR
CA
CL
CZ
DK
EE
FI
FR
DE
GR
HK
HU
IN
ID
IE
IL
IT
JP
LU
MY
MX
NL
NO
PK
PL
PT
SG
KR
ZA
ES
LK
SE
CH
TW
TH
TR
GB
US
Name of Index
Merval
ATX
ASX ALL ORDINARY
Bel20
BOVESPA
S&P/TSX Composite
IGPA General
PX50
KFX index
Talse index
HEX General Index
CAC40
DAX 30 Performance
ASE General
Hang Seng Index
Budapest index
Bombay SE 200
Jakarta Composite Index
Overall index
Tel Aviv SE Moaf 25
MIB 30
TOPIX
Datastream market index
Kuala Lumpur Composite Index
IPC (Bolsa)
AEX
OBX
Karachi SE 100
WIG
PSI (BVL) General
STI
KOSPI
JSE All Share
IBEX 35
Colombo SE All Share
Affarsvarlden General Index
Swiss Market Index
TWSE
SET Index
ISE National 100
FTSE 100
S&P 500
Journal of Applied Business and Economics vol. 14(3) 2013
Datastream Mnemonic
ARGMERV
ATXINDX
ASXAORD
BGBEL20
BRBOVES
TTOCOMP
IGPAGEN
CZPX50I
DKKFXIN
ESTALSE
HEXINDX
FRCAC40
DAXINDX
GRAGENL
HNGKNGI
BUXINDX
IBOM200
JAKCOMP
ISEQUIT
ISTMAOF
ITMIB30
TOKYOSE
TOTMKLX
KLPCOMP
MXIPC35
NLALAEX
OSLOOBX
PKSE100
POLWIGI
POPSIGN
SNGPORI
KORCOMP
JSEOVER
IBEX35I
SRALLSH
AFFGENL
SWISSMI
TAIWGHT
BNGKSET
TRKISTB
FTSE100
S&PCOMP